Get your ticket
business triangle technical hexagon

Generative models (Gaussian Mixture Models) for images and time series

Technical talk | English

Track 4 - Theatre 15

Wednesday - 17.50 to 18.30 - Technical

Tags:
- -

Generative models (Gaussian Mixture Models) for images and time series.

This Session guide user to deep dive into generative models. The goal is to use a Jupyter notebook and data from the time-series and image dataset to illustrate various techniques for generative models.

Gaussian mixture models (GMM) are a probabilistic model for representing normally distributed subpopulations within an overall population. Mixture models in general don’t require knowing which subpopulation a data point belongs to, allowing the model to learn the subpopulations automatically in unsupervised learning settings.

In this talk, we will deep dive into intuition and working of GMM. We will cover many use cases:
Using GMM for predicting time series like stock prices .
Using GMM as generative model for images i.e a model can generate a new face who doesn’t exist.
Using variant of GMM, we will do object detection.

As a data-scientist, it is very common to work with different models and especially with the generative model which has far reaching implications that non-generative models.

In this session, user will learn end to end techniques for various use cases in time series, images and object detection.

Here are some of the contents in as pdf/notebook.
https://github.com/aloknsingh/ds_deepdive_gmm/blob/master/doc/pdf/GMM.pdf
https://github.com/aloknsingh/ds_deepdive_kmeans/blob/master/doc/pdf/deep_dive_kmeans.pdf
https://github.com/aloknsingh/ds_deepdive_gmm/blob/master/notebooks/GMM.ipynb
https://github.com/aloknsingh/ds_deepdive_kmeans/blob/master/notebooks/deep_dive_KMeans.ipynb

Sponsors

FIDELITY
ORGANIZER
KUDOS FOR